Abstract
Tillering plays an important role in determining rice grain yield. Several models have been developed to predict tiller production in irrigated rice. In this paper we tested three models using data drawn from a wide range of plant densities and N inputs in two field experiments conducted at the International Rice Research Institute, Philippines during the 1997 and 1998 dry seasons. Two rice cultivars (IR64 and IR72) were used in the experiments. Plant samples were taken at intervals to determine number of tillers, leaf area index (LAI), biomass, relative growth rate (RGR) and leaf N concentration. The models were parameterized using an iteration procedure of the simplex method. Previous models (TIL and SINK) using the original values of parameters for IR64 failed to predict the number of tillers of IR64 in 1997. However, when re-parameterized, both models described the 1997 data well for both cultivars. The two models also predicted fairly well the number of tillers of different transplanting spacing and N input treatments of IR72 in 1998 using IR72 parameters derived from the 1997 experiment. A simple RGR model was comparable with the TIL and SINK models in descriptive and predictive ability. It appears that all three models could be used for predicting tiller production of irrigated rice.
